topic Proc Reg analyses Stratified by categorical variable in Statistical Procedures

Proc Reg analyses Stratified by categorical variable

battsetseg — Wed, 29 Jan 2020 23:12:50 GMT

Hello,

I would like to run the proc reg procedure stratified by some categorical variables.

this is what I did without stratification:

Proc reg data=DATA; Model DEPENDENT=independent_1 ; run;

Now, I would like to calculate the estimated effect of the variable "independent_1" on DEPENDENT but stratified by Independent_2. "Independent_2" variable has four categories (0, 1, 2, 3). I would like to see for Independent_2=0 then what is the estimated effect of the variable "independent_1" on DEPENDENT.

Thank you

Batts

Re: Proc Reg analyses Stratified by categorical variable

PaigeMiller — Wed, 29 Jan 2020 23:24:17 GMT

You can create regressions by categorical variable with the same slope across each level of the categorical variable, or you can four regressions with different slopes.

proc glm data=data;
    class independent_2;
    model dependent = independent_1 independent_2
        independent_1*independent_2
        /solution;
run;

This gives four different slopes. If you want them to have the same slopes, remove independent_1*independent_2

Re: Proc Reg analyses Stratified by categorical variable

battsetseg — Wed, 29 Jan 2020 23:45:04 GMT

PaigeMiller,

Thank you for the suggestion. I run my code but it is not showing any estimate effects. So I would like to see the estimated effect of "Independent_1" on "Dependent" by if each category/level of "independent_2" (this has for levels 0,1,2,3).
For example: When "independent_2" = 0 what is the estimate effect of "Independent_1" on "Dependent".

When "independent_2" = 1 what is the estimate effect of "Independent_1" on "Dependent".

For example: When "independent_2" = 2 what is the estimate effect of "Independent_1" on "Dependent".

For example: When "independent_2" = 3 what is the estimate effect of "Independent_1" on "Dependent".

I would like to see the estimate effect, CI, P-value, and Number of observations used in each of these results.

thank you very much!

Re: Proc Reg analyses Stratified by categorical variable

PaigeMiller — Thu, 30 Jan 2020 12:04:12 GMT

Did you use the /SOLUTION option? If so, the estimates should be shown in the output.

Please show us the code you used. (Click on the running man icon and paste the code into that window)

Please show us the output you obtained.